Abstract
No-reference bitstream-layer point cloud quality assessment (PCQA) can be deployed without full decoding at any network node to achieve real-time quality monitoring. In this work, we develop the first PCQA model dedicated to Trisoup-Lifting encoded 3D point clouds by analyzing bitstreams without full decoding. Specifically, we investigate the relationship among texture bitrate per point (TBPP), texture complexity (TC) and texture quantization parameter (TQP) while geometry encoding is lossless. Subsequently, we estimate TC by utilizing TQP and TBPP. Then, we establish a texture distortion evaluation model based on TC, TBPP and TQP. Ultimately, by integrating this texture distortion model with a geometry attenuation factor, a function of trisoupNodeSizeLog2 (tNSL), we acquire a comprehensive NR bitstream-layer PCQA model named streamPCQ-TL. In addition, this work establishes a database named WPC6.0, the first and largest PCQA database dedicated to Trisoup-Lifting encoding mode, encompassing 400 distorted point clouds with both 4 geometric multiplied by 5 texture distortion levels. Experiment results on M-PCCD, ICIP2020 and the proposed WPC6.0 database suggest that the proposed streamPCQ-TL model exhibits robust and notable performance in contrast to existing advanced PCQA metrics, particularly in terms of computational cost.
Introduction
As 3D point clouds become the backbone of applications ranging from autonomous driving to digital heritage, the computational cost of processing these massive datasets has become a significant bottleneck. Compression standards like G-PCC (Geometry-based Point Cloud Compression) are essential for storage and transmission, but they introduce perceptual distortions. Monitoring this quality in real-time has traditionally required full decoding of the point cloud, a process that is too slow for network-level monitoring. This paper introduces streamPCQ-TL, a no-reference bitstream-layer model that predicts quality by analyzing the "DNA" of the file without full decoding.
Methodology: The streamPCQ-TL Model
The streamPCQ-TL model adopts a physics-informed approach to quality assessment by modeling the relationship between three key bitstream parameters: the Texture Quantization Parameter (TQP), Texture Bitrate Per Point (TBPP), and the geometry parameter trisoupNodeSizeLog2 (tNSL).
A core innovation of this work is the estimation of Texture Complexity (TC). The researchers discovered that at a constant quantization level (TQP), there is a nearly linear relationship between texture complexity and its bitrate. By leveraging this, they could estimate TC without decoding the geometry. The final model combines a texture distortion model with a "geometry attenuation factor" derived from the tNSL parameter, reflecting the dissociable nature of texture and geometry distortions in Trisoup-Lifting encoding.
The WPC6.0 Database
To validate the model, the authors established WPC6.0, the first and largest database dedicated specifically to Trisoup-Lifting. It contains 400 distorted point clouds derived from 20 high-quality sources, covering a wide range of geometry and texture distortion levels to ensure robust training and testing.
Experimental Results
Experimental results on the WPC6.0 database show that streamPCQ-TL achieved a Pearson Linear Correlation Coefficient (PLCC) of 0.9562 and a Spearman Rank Correlation Coefficient (SRCC) of 0.8589, outperforming several full-reference models that require access to the original uncompressed data.
The most significant advantage is computational efficiency. For a point cloud with approximately 2.8 million points, a standard spatial-layer model like GraphSIM requires 653 seconds to run. In contrast, streamPCQ-TL completes the assessment in just 0.023 seconds—a 28,000x speedup. This performance shift enables instantaneous quality monitoring at network nodes.
Conclusion
The streamPCQ-TL model provides a highly efficient and accurate solution for real-time quality monitoring of 3D point clouds. By operating at the bitstream layer, it avoids the heavy computational lifting of 3D rendering and spatial analysis. The public release of the WPC6.0 database and source code at https://github.com/qdushl/Waterloo-Point-Cloud-Database-6.0 will further support the development of lightweight 3D quality metrics.
Summary updated